<p>In current optical image encryption schemes based on human biometrics, most approaches rely on physiological biometrics. Moreover, these schemes are typically designed for single images, and the encryption results are often noise-like images. To address these issues, this paper proposes a visually meaningful encryption scheme for dual color images based on chaotic voice phase masks. In our proposed encryption scheme, the human behavioral biometrics, i.e. the human voice, can directly replace the random phase masks as the secret key in the encryption and decryption process. This substitution simplifies the management of the random phase masks and eliminates the need for their transmission. Furthermore, our proposed encryption scheme enables the encryption of two color images into a single visually meaningful image, ensuring that the encrypted result does not attract the attention of potential attackers during network transmission. Finally, extensive numerical simulations have also confirmed the feasibility, security, and robustness of our proposed encryption scheme.</p>